banding... more complex the longer i work on the t

[tomasz.k]

Hi all,

I'm new on the forum, so sorry if this topic is covered in full somewhere but I did look around and did not find a real answer.

We often have problems with banding in our studio. I'm sure we're not the only ones around... Sometimes we manage to cope with it quite quickly and easily, but there were times when we really were desperate for better methods as the deadline was mercilessly close and we really couldn't come up with a proper solution.

Underneath I will try to list the methods I usually try to use to cope with the problem as well as a few thoughts on the nature of the problem. I really do hope that there is whole bunch of people here who will efortlessly solve my headache;)


1. Upon seeing banding on screen I usually first look through seperate RGB and later Lab channels to see how it is constructed.

2. If it happens to be simple color banding (which means it isn't visible on the L channel but just on ab in Lab) then of course removing it is a snap. Solutions would be numerous: paint over with a color layer, blur sections of a and b channels....

3. If this is not the case, if I have been working on an 8bit file and have the raw, then I try to redevelop in 16 bit and see if this helps. Rarely it is a solution, but it helps a bit in may cases. (And often it is a question of small steps to get rid of it completely.)

4. Lets assume that 16bit developing didn't help and I'm stuck with my 8bit file. Also that the photographer does not want to alter the image look so I can not just redo the background by making a new gradient with a different graininess and color. What next then:

5. All of the steps underneath should work better if done in 16bit, at least in theory;)

6. Sometimes if it is just one place where the uneven transition is visible painting on a color layer helps to at least partially get rid of the effect. Then we can for instance even out the difference in luminosity by using dodge and burn lets say through a softlight layer.

7. Another trick is to (of course) add grain to that section and then reduce it through median or reduce noise on seperate layer which is a copy of the section, place it on top and set it as a luminosity layer, then duplicate it and set it as a color layer. Then reduce opacity of the layers to the desired effect, mask edges or fragments....

8. Copying the banding section and taking it to Lab and blurring the color channels and then evening out the L channel with dodge and burn and taking back to RGB and placing on top and then setting this at the proper opacity also often helps.

9. If the problem is clearly visible on one or more fo the seperate RGb channels sometime delicate dodge and burn, healing brush (etc) operations help.

10. If it is not affecting large areas sometimes just healing the images in the areas affected can break up the lines...

11. Creating a color layer and healing, stamping from nearby areas to create an even tone and then if it is also a luminosity problem healing on a luminosity layer can work

12. Making tons of circular gradients of tone very close or even picked from the affected place, and using them on very low opacity creates a gradient with a structure so complex that banding line don't appear. Sometimes this helped me a lot.

13. If things get really bad I do sometimes try to just create that section from scrap and then recreate the grain or structure of the image, but this is not really a solution. It's just a way around the problem which in most cases would use up too much time or just be impossible.


What is really bothering me though is that the banding we see on screen is sometimes not visible on the actual output file. We recently did a cover for a magazine; the image was "almost" bw, in fact it was a RGB but developed with ) saturation but with channel clipping in Capture One(channel B was taken down to around 245 from what I remember, which gave the image a nice yellowish tone) which resulted in very subtle color variantions. The file was already developed with the clipping so it was not a question of 8bit curve manipulation in ps. In the shadows there were visible green banding lines.

Now THIS is what bothers me: the photographer saw it in different places then we did. We changed the color to an even tone in the whole section and it was still visible there but the look changed (the color picker suggested though that our monitors are fooling us). We began to suspect that this is a software/ hardware rendering problem so we proofed it. There was no banding there.... My theory is that sometimes banding shows in places were it actually is not present, just the software/ hardware rendering shows it due to inadequate hardware color rendering possibilities.

Still it does not change the fact that even if the banding is not physically there a retoucher has to remove the appearence of it to make the photographer happy;)

Sorry for the long post. What do you think?
Tomasz

tk@houseofretouching.com
www.houseofretouching.com

[TommyO]

Tomasz,
Welcome to RetouchPro ! Sorry about your problem, it sounds very irritating.

It also sounds like a mixture of three issues.
(a) low bit depth; (b) broad color space - hungry for more bits; (c) operating system bugs.

A few questions:
What color profiles are you using ?
Are you working in a Mac environment ?
Are you up to date on OS patches ?
Are your video drivers up to date ?
And your client/photographer... same questions.

The MAC OS was having similar issues in 2005-2006. Since then many of those have been patched up. The fact that you are not seeing it in proof also makes me lean this way.

[tomasz.k]

Hello Tommy,
Glad to see you have some clues although I do not really think this could be an OS issue, but I am more than open to such a possibility.

Do you use any other methods for removing banding though?

Answers underneath:

What color profiles are you using ?
We work in Adobe 1998, monitors calibrated with Eye One
Are you working in a Mac environment ?
Yes Leopard
Are you up to date on OS patches ?
up to date with everything
Are your video drivers up to date ?
Not sure, but we see the same thing on a G5 and MacPro, so I don't think it's a video card driver problem
And your client/photographer... same questions.
He was working on a Macbook Pro with an external monitor, don't know what kind though


I also have one more possible solution, could it be that we are still using LCD's which have 12bit hardware and not 16bit? We still use NEC MultiSync® LCD2190UXi monitors at the moment, but want to upgrade to Eizo CG series when the opportunity arises.

I think when I will have a moment tomorrow I'll crop that part of the cover and post a link to see if you see banding in the same spot.
Cheers
Tomasz

[HuBBa]

I would personally say this is most likely hardware related, especially since you say that you see banding in different areas than the photographer/client do on their equipment.

So to avoid it, get as high quality gear as you possibly can. If budget don't allow for the better Eizo screens, use prints (from a GOOD printer) as references. Its slower but generally the end media where you bother with banding issues is print anyway. Web and banding issues is like worrying about hi-fi sound mixing on a mp3 player

[tomasz.k]

Originally Posted by HuBBa

I would personally say this is most likely hardware related, especially since you say that you see banding in different areas than the photographer/client do on their equipment.

So to avoid it, get as high quality gear as you possibly can. If budget don't allow for the better Eizo screens, use prints (from a GOOD printer) as references. Its slower but generally the end media where you bother with banding issues is print anyway. Web and banding issues is like worrying about hi-fi sound mixing on a mp3 player

Yes that is true, but this was just one occurrence. Usually we see banding in the same areas and more then often it does come out in print.
My general question was though do you have any other methods of dealing with the banding?

[HuBBa]

Ah sorry. should have written a bit more.

If the banding shows up in print, then the dynamics have been compressed somehow. This usually happens when using certain tools like curves which re-shapes the range of tones. Its as you have noticed already very hard to avoid, and re-painting in a smooth gradiant can be pure hell.

I'm personally no expert on which tools compresses tonal ranges but if im not mistaken, the obvious ones are curves, levels, and hue/saturation. So avoid using those on smooth gradient areas as banding will most likely occur.

What im not however 100% sure is how much compression is done if curves/contrast adjustments are done on the RAW original before going into normal pixel-manipulation.

This is discussed a bit in a video called Adobe Photoshop Cosmetic Techniques (http://www.thestudio2u.com/etntworld) but there ought to be some info about this on the net aswell. It's mostly evident with tools that affect the entire tonal ranges ofcourse, so colorbalance, hue/sat, curves, levels, etc. and blendmodes will effect this.

The way to avoid it (as i see it as a photographer) is to start of with as perfect exposure as possible, which will need as minimal tweaking afterwards. Since once you start messing with tonal ranges, banding may occur. And ofcourse going with as high of a bitdepth as you can in the original files.

Sorry if this isnt much help to what you already know but its a problem i battle with aswell.

[tomasz.k]

I totally agree with you. We do try to do as much as possible through raw manipulation. Still photographers often ask us to push the raw VERY far, so although we develop to 16bit we end up with banding from the beginning...

[stopa]

This is the worst nightmare for us... And I think that it is our faith... Since Newton rings appears, on the first enlarged prints.

first of all - Thanks for your post. Now I know that not only I fight with this - and it is not only my paranoia. I thought, maybe I don't know something and I have some holes in my knowledge?

Tomasz- why you work on LAB? Every time you convert to other space (RGB or CMYK) all gradients are re-counted - and ALWAYS bad things happen to all gradients.
(you can trap final step as smart object - then convert it)

Yours banding - more known as Moire, or Tonal pass - it is a matter of bit depth.
And I will try to illustrate it for others.
We have 256 steps on 8 bit for each channel. it is far enough when you have color-full image, like parrot sitting on a triangle. But we often work with semi-tonal images. Almost white skin, and very dark - black - background.
Now you know that you have not AS much space! Only well balanced colours are neutral. So you now have only few bit left to work on!
And now we are going further: Background
You have 5000(+) pixels high image, but you NEED totally smooth gradient - but wait... DARK - gradient! Now you have only 15-50 steps for this!
(take a sample for lightness on background - "L" indicate values from 0 to 30...)
Once again: it is simple arithmetic 5000 pixels in 50 steps - each step is 100 pixel high! What happens when you have only 5-to-15 background lightness? And mathematics never fails... sorry...

My solution is dodge and burn, and blur and noise.

Don't even think that 16bits will help you out with this... CMYK is only 6bit or even less (0-100% of ink for each channel)... and our work is often finally converted for it.
Dan Margulis said - that it is essential to build gradients in the final color mode.

- Tell as more about - if your bands was shown on proof, or on display?
- I may say that apple displays are waste of money (you can pass it to your client)
- Show this image to us, I am really curious.

And finally - welcome to RetouchPro!
Fajnie że się spotykamy właśnie tu na retouchPro Pozdrawiam W

[tomasz.k]

Originally Posted by stopa

Tomasz- why you work on LAB? Every time you convert to other space (RGB or CMYK) all gradients are re-counted - and ALWAYS bad things happen to all gradients.
(you can trap final step as smart object - then convert it)

Yours banding - more known as Moire, or Tonal pass - it is a matter of bit depth.
My solution is dodge and burn, and blur and noise.

Don't even think that 16bits will help you out with this...

I convert parts of the image or the whole thing to LAB because it is the easiest color space to solve problems in. As least a large part of them
I have never seen the operation of going from RGB to LAB and back make the problem worse so I don't worry about the small disturbance it creates. To say the truth LAB usually is of great help to me.

I think I am correct though on the term- banding, it is different from moire in my understanding. Moire we usually get on clothes were patterns don't go well with the pixel structure of the image. Moire we cope with quite well, especially that it usually happens on small areas.

16bit does sometimes help. We recently did a series of shots for reebok on white backgrounds. On some of the 8 bit files we began to get very delicate banding, which practically disappeared after redeveloping to 16bit.

[stopa]

Originally Posted by tomasz.k

I convert parts of the image or the whole thing to LAB because it is the easiest color space to solve problems in. As least a large part of them

to be honest I prefer Lab also

Originally Posted by tomasz.k

I have never seen the operation of going from RGB to LAB and back make the problem worse so I don't worry about the small disturbance it creates.

It occurs on wide gamut images.

Originally Posted by tomasz.k

I think I am correct though on the term- banding, it is different from moire in my understanding. Moire we usually get on clothes were patterns don't go well with the pixel structure of the image. Moire we cope with quite well, especially that it usually happens on small areas.

Yeap, you are right, but very often common people name this kind of bands a moire, and don't be startled when your client will say: "Hey, there is moire!" , and you will not understand what is he/she talking about.

Originally Posted by tomasz.k

16bit does sometimes help. We recently did a series of shots for reebok on white backgrounds. On some of the 8 bit files we began to get very delicate banding, which practically disappeared after redeveloping to 16bit.

Every case has its own solution.
Those images was finally printed or what?

[HuBBa]

And just to clarify for those readers not familiar with the terms, banding is when you see distinct "bands" in tonal transitions showing a lower tonal resolution in areas.

Moire is a interference pattern created by very small repetetive patterns (such as fabric texture, prints, etc) which creates a wave pattern when digitized. This is a HUGE simplification of the actual process going on but for ease of explanation i think it suffices.

And you can ofcourse experience both at once

As for clients demanding you push the RAW a lot, hit them hard with a mallet As a photographer who has to do his own retouching, if i have to push the RAW to its limits, it means i haven't personally done my job. And whenever i have to start pushing things in the RAW developer or in photoshop to save an image, thats when things will not be as good as they should have been.

My personal preference as a photographer is to think like i was shooting film (with the added knowledge of how digital sensors differ from film emulsion =) in the sense that the image should be 99% done in camera. This means makeup should be spot on, clothes & styling & model should be spot on and the exposure and lighting should be spot on.

This means that the stuff i then do in Photoshop, is enhancing the image instead of fixing mistakes i should have taken care of before i pressed the shutterbutton. And when you have to fix mistake, you pretty much always have to degrade the image in some way (tonal ranges, dynamic range, textures, resolution, etc) which means you end up with a lesser quality image than you started with even if it may look more like you wanted. But it will most likely print worse.

The old "don't worry, he'll fix that in photoshop" earns an earlashing in my studio

[tomasz.k]

Originally Posted by stopa

Those images was finally printed or what?

to be printed anywhere they want to we usually retouch not knowing where the image will end up, as it is rarely one media or publication

[stopa]

I can't agree with Hubba! Well done post production can tease out the other dimensions from images. And as more you know about it, you will shoot building in to your workshop a postproduction step.

Eye of camera is totally impartial. But you - as a photographer, as human, as artist you are not!
I can agree that PRIMAL thing it is to make a good photo - and retouch is only add-on.
But proretouchers never degrade the image.

------------------

My general question was though do you have any other methods of dealing with the banding?

- I was using very heavy motion blur (more than 200) last time on a gray background (the model was path-out from the background), and than add a image grain for it. looks fine.
- I hate vignetting, I build gradients in destination mode. In print looks fine.

And remember: Proof is a Proof - if you notice something on display but nothing in proof print, do not bother. How it will look on a paper it is always a Russian Roulette, and until you are not a printer (pressman), you are not responded for it. I charge for every size, format, and print process. (image is still the same)

[tomasz.k]

Originally Posted by HuBBa

As for clients demanding you push the RAW a lot, hit them hard with a mallet
...
My personal preference as a photographer is to think like i was shooting film
...
The old "don't worry, he'll fix that in photoshop" earns an earlashing in my studio

i totally agree. How many times have I thought that if they tried a touch harder I could spend those hours making them even better instead of cleaning up the mistakes.

Still there are real photographers out there. Some people we work with really shoot amazingly good stuff which is good from the moment of development.

Still banding is an issue they can't plan during shooting, often cutting off channels to create an interesting look creates it... and that's where the topic of my post comes back.

[stopa]

i totally agree / I can't agree with Hubba!
funny!!!

[HuBBa]

Stopa, what i meant was a bit different though. If you exposure correctly for the intended post production, thats a whole different ballgame than if you have to push up exposure 2 stops becuase the photographer didn't set his lights correctly to get detail in the shadows.

Again... Fixing ERRORS in post is a bad idea. Shooting to enhance your images in post is a good idea. =)

[SilvaFox]

Some very good suggestions. And obviously you're not alone in this. We ALL have had this problem one time or another.

I can tell you this. The more you mess with it, the worse it gets. If you get to the Curves and Brushes stage, you're chasing your tail.

I find that slightly blurred noise, blended over the breaks with layer masks, then set to multiply is the best way to deal with it. We apply it in the colors of the background only, so it doesn't dirty everything up too much. (C,M noise only in a sky for example.) As you know, the key is to not yank everything all over the place from the beginning, creating it, then doing the same thing trying to fix it.

I empathize. It's difficult to sell the "It's there but it won't print" theory.

Good luck.

[neumanns]

As an novice I am following this thread. I should just keep my head down and let you pro's have at it, but I'm gonna open my mouth and insert my foot anyways.

As for the solution I don't know. Seem's like you have quite a few workarounds in your toolbox.

But it appears to me the problem is quite simply rounding. When you push and pull the pixal's the value is not infinite so when the computer renders the image it has to round the value to a finite number (specified range).

This explains why it appears on diffrent locations on diffrent machines. Many programs and hardware handle the math slightly diffrently.

You may just have to accept that you are at the limits of what is capable with the data you have. And continue with the "fix it" approach.

[tomasz.k]

SilvaFox- very nice idea, I'll definitely give it a try

Neumanns- yes we eventually always have to fix it)

[TommyO]

Originally Posted by tomasz.k

Hi all,
...the photographer saw it in different places then we did.

...even if the banding is not physically there a retoucher has to remove the appearance of it to make the photographer happy.

Tomask,
I am thinking that the problem is a combination of the larger working color space (AdobeRGB), damage to the histogram due to various edits, low final image bit depth (8-bit), and somewhat low monitor bit depth (12 bit internal LUT). So, it may simply be unavoidable unless you change one of the above mentioned elements.

Since you mentioned it is seen in different areas at different times, I cannot agree with your second point, having to retouch it out. You're just chasing a moving target, accomplishing nothing if the posterization is going to move again.

The histogram is the only tool you have at your disposal to really judge if posterization is present in the image. If the histogram is smooth (as is each channel of the histogram), then it's not in the image and your edits are not the issue.

It also sounds like some testing is in order. Create an image with subtle gradations in AdobeRGB. (Just download a standard output profile.) Save output files with graduated editing problems and check on your hardware. If the posterization varies in location and is not reproducible, then it is your hardware.

One workaround is to convert to a smaller color space for final soft proofing. Reduced gamuts in low bit environments can result in less posterization in the output, whether on the monitor or printer.

Oh, and be sure your calibration workflow is working correctly, calibrating the internal LUT of the monitor and creating a linear (nothing) profile for the graphic card LUT. Just in case.

[pixelzombie]

can you post a portion of the file that appears to have banding on the screen?

[tomasz.k]

Here is a link to the photo I was referring to. It's just a part of the background, but I can't show any more unfortunately, it hasn't been published yet.

On the right hand side I drew the areas where I see green and magenta

http://www.houseofretouching.com/Download/banding.tif

and here is a link for the file just after developing
has even worse banding at least on my screen

http://www.houseofretouching.com/Dow...anding_dev.tif

Hope you see the same... I checked to make sure and our monitors have 12bit LUT and 12 bit internal processing, so upgrading to Eizo CG should make the difference you were talking about Tommy.

[tomasz.k]

Originally Posted by TommyO

I am thinking that the problem is a combination of the larger working color space (AdobeRGB), damage to the histogram due to various edits, low final image bit depth (8-bit), and somewhat low monitor bit depth (12 bit internal LUT). So, it may simply be unavoidable unless you change one of the above mentioned elements.

The thing is though, that we develop to Adobe 1998 from C1 and try not to do too much curve work in PS. So if what comes out of C1 has banding on an average screen I guess we have to retouch it somehow even if it's not there. After all the image will surely be printed but it usually will also end up in the web at some point and none of our photographers want to hear the "it will look good in print" theory

Originally Posted by TommyO

Since you mentioned it is seen in different areas at different times, I cannot agree with your second point, having to retouch it out. You're just chasing a moving target, accomplishing nothing if the posterization is going to move again.

Yes, but again what can we say to the client in that case... but it is definately a point that we ought to be able to tell weather what we see on screen will print that way or is just a case of our hardware misleading display, so I guess we will have to upgrade very soon from our 12bit LUT and 12 bit processing Necs to Eizo CG series, but they also have 12 bit LUT, it's the processing they have at 16bit

Originally Posted by TommyO

It also sounds like some testing is in order. Create an image with subtle gradations in AdobeRGB. (Just download a standard output profile.) Save output files with graduated editing problems and check on your hardware.

Sorry Tommy I don't quite follow this part?

[tomasz.k]

I decided to check both of these links at home on a 4 year old imac which hasn't been calibrated for half a year (so I would say it's a standard users computer)and ... I can see less banding here then on our much more expensive studio monitors. It is there though in the same places, but much less saturated.

[stopa]

i can see it even on my old CRT

[TommyO]

Tomasz,
I'm starting to think it has little to do with hardware and a lot to do with variations in human vision.

I see no posterization at all. However, I do see subtle variations in color and tonality that simply look like intentional variation in the background. The variations are very tricky to the eye, as well as to the LCD display and lighting conditions.

My eyes are very evenly sensitive to all three primary colors. Other peoples' do vary a great deal. Anyone more sensitive to red or green will certainly see more posterization due to the component colors in your particular background. (See attached thumbnail)

Since the variations are very subtle, low bit hardware could certainly enhance it and be a factor in the degree in which your eyes perceive it.

I did see some clipping in the shadows of the clothing, as well as in the blues. But, that should not affect the background posterization issue.

In the end, I think you have chosen a background color that is tricky. Some people will see it as a soft blend, similar to a brownish, greenish, gray color. Others are going to see posterization on those zones where the reds and greens shift, simply due to their own eyes being more sensitive to those colors. On the thumbnail, I chose a 3x3 average since most adjacent pixels were very similar in value. However, while my eyes can see the hue shift from just a few % in the next zone, I don't perceive it as posterization/banding.... just a subtle shift. It would be interesting to see what a sample of 10 or more people see on your monitor, with no hint of what they should expect to see.

[pixelzombie]

i looked at both images on apple and eizo monitors and that's not what i would call banding..i see that as being digital artifacts from the cameras sensor, it also happens with film images that are scanned on low end drum scanners..the dynamic range is just too high for the sensor to record accurately if that makes sense...

[stopa]

Notice that I have foresee that tomasz.k have problem with semi-black background. It is common thing, and as I said before it is a raw arithmetic.
Tomasz.K work in conformity with the regulations, this is what you can achieve in 8bit scale.
And I would be surprised if this thing will be perceptible on A4+ coated cover.
Unfortunately from my experience I can said, that the simplest things are as heaviest (time consuming) as possible!

-I think that you can add some noise, if you wish.

[neumanns]

I am on a low grade monitor, But I do not see banding. What I see is color variation in groups if that makes any sense. To me banding is uniform.

Do you get the files completly untouched. Straight from camera with no styles, contrast, sharpening, Etc.... From looking at the undeveloped tiff there seems to be some contrast issues. I only shoot a canon 40D but if you look at 100% view of the clothing there seems to be some extreme variations from neighboring pixals in areas where there should be a softer transition.(lighting alone should not cause this abrupt of change)

My guess has changed...I think the problem stems from post processing before conversion to tiff. Possable causes could be curves, contrast, or "in camera" style settings.

I would be surprised if a camera raw file with no "presets" showed the problem as prominent.

[tomasz.k]

I'm back. Too much to do today to answer earlier.
So you think it's not banding, do you? I must admit that a lot of the files we work on are developed in C1 with channel work on them (and cutting off channels often creates a mess, but the effects look astonishing in my opinion). Some of these "color variations" (for me still banding) might be very subtle but they are obviously faults we have to mend in the retouching stage.

More then once we experimented weather doing the correction in PS would not give lesser banding, but that road always fails... and I'm happy about it as I do prefer to think of the RAW as the negative out of which I have to push out as much as possible before altering it in ps.

and so thank you all for the help. From these posts I have learned a lot and if I come up with a different solution or understanding of the problem will definitely renew the thread. I'm very curious weather our hardware upgrade in the future will free us of these problems.